kumoh national institute of technology
Networked Systems Lab.

Review

NSL > Education> Review
Long- [IEEE Transaction on Industrial Informatics]
By :
Date : 2017-02-09
Views : 60

Encl.:
Reviewer: 1

Comments to the Author
The paper proposes a new routing protocol for industrial IoT networks. The proposed scheme aims at reducing energy consumption of the nodes. Presented simulation results show that the protocol outperforms previously proposed solutions. The paper is well-written, it is easy to follow and understand. However, a set of things has to be improved before it can be accepted.

Research goal and contributions:

First, the research goal of the paper has to be explained better. It is stated that the goal of the work is to develop an energy-aware routing protocol for industrial IoT applications, but more motivation is needed. In general, in industrial applications the main requirements to communication systems are reliability, availability and timeliness. Current paper looks at the possibilities of energy consumption decrease instead and thus more explanations and examples of particular applications in mind are needed. Next, it is said that the paper focuses on IoT and not IWSANs as IoT covers larger areas and number of devices. But it has to be explained more clearly why protocols developed for IWSANs cannot be scaled to IoT and why new protocols are needed. And, given that, contributions of the paper and the differences between ERRS and previously proposed protocols have to be explained better. As it is stated in the paper, cluster selection algorithm used is very similar to the ones from references 1, 7 and 13, while the routing scheme is close to 21. The new features added in this work have to be highlighted better.

Protocol design:

Due to periodic nature of transmitted packets and to guarantee deterministic delays, TDMA is often used in industry (and in wireless standards developed for industry, such as WirelessHART, ISA 100.11a and WIA-PA), while the current paper adopts CSMA/CA. This choice has to be explained since CSMA cannot guaranty predictable channel access delays, which means that deadlines can be missed and industrial requirements on timeliness might not be fulfilled.

Moreover, it is said in the paper that CH collects the readings from several sensors before one aggregated packet is sent to the gateway. This will save energy spent on packet transmissions, but will increase latency for packets that arrive to CH early and have to wait before all the other data is collected. This trade-off between energy and time has to be discusses as well, as in industrial systems requirement on real-time communication is more important than on energy saving.

Transmission order in the routing scheme should be explained more clearly. It is said that when one CH sends RREQ, its neighbors reply with a message including their hop count to the GW. To do so they should know the hop count. This means that the GW should start this initiation process, or? And the last sentence in Section II.B is confusing. Algorithm 2 is followed only once in the initiation stage, not for every data packet, or?

Also, only one path is selected by the proposed protocol, while industrial systems often require redundant paths to ensure systems availability in case of failures (e.g., as stated in [1], WirelessHART requires every sensor node to have at least two available paths towards the gateway). How will the protocol behave in case of a link or node failure? Will the network operation have to stop to establish new routes? This has to be discussed in the paper since systems reliability and availability are the key requirements in industrial systems.

Performance evaluation:

Moving to the results section, simulation model should be explained better. No scheduling is mentioned anywhere. It is said that different clusters can be allocated different channels, but no reference for suitable channel allocation scheme is given. And during the set-up phase clusters are not known yet and thus nodes will interfere with each other when trying to create the clusters.

Moreover, simulation setup has to be explained in more details. How and how many data packets were sent? How many transmission rounds were simulated? How many different node placements were tested and do the presented results show the data for one placement or an average for several placements? Comparing different placements, do they all show the same performance or are there any unlucky ones (e.g., since the nodes are placed randomly, is it possible that there are I/O devices that are located too far away to hear any router?)? Which exactly channel model was used and were channel conditions stable during the simulations? How would the protocols behave if channel conditions were changing over time (which is often the case in harsh industrial environments)?

Next, the proposed protocol is compared with three existing ones. References to the protocols are given, but it would be good to add short descriptions as well, so that the reader is not forced to go to the mentioned papers. And it is especially important to describe the differences since it is stated that the proposed scheme is very similar to the one from reference 21. Also, on page 8 it is said that scheme in 21 supports only single-hop communication. Is it so? And if yes, then maybe it would be better to compare the proposed scheme with protocols supporting multi-hop?

Simulations were done only for one size of the operating area. What if the nodes were placed in a smaller area, e.g., 100*100 m, would the compared schemes perform in the same way? If the benefit of ERRS is better performance in large-scale deployments, it would be interesting to see the borders of where it starts and stops outperforming the other schemes.

Also, it is shown that the proposed protocols outperforms the other ones in terms of energy consumption, but does it provide the same reliability? It would be interesting to see that, e.g., packet reception rate in not degraded due to different route selection scheme.

And it would be good to add more explanations to the figures with the results, e.g. why energy consumption decreases with time in Fig. 8? Why do so many nodes run out of battery after only 900 seconds of simulation in Fig. 10? In industry the devices typically work 24 / 7 and it is not possible to change batteries too often.

Overall paper comments:

Description of IEEE 802.15.4 is spread between System model and Results sections, which might be confusing for the readers. It is good to have simulation parameters in the Results section, but general overview of the standard would fit better in Section II.C (which should probably be renamed to IEEE 802.15.4a overview, since it is stated that 802.15.4a and not pure 802.15.4 was used).

The paper has a lot of figures and algorithms which makes it very easy to follow. However, it exceeds TII page limit and has to be shortened. One way to do it could be to remove some of the figures and algorithms.
- For example, Fig. 2 can be removed since all three topologies are most probably known to the readers, star and mesh topologies are not mentioned anywhere else in the paper and cluster-tree can be explained using Fig. 1 if needed.
- Next, Fig. 3 and 6 are very similar to each other. It would be fine to keep only one of them or remove both since they are well described in the text.
- Fig. 4 and 6 could be removed as well as the text explained processes well. If they are kept, they should be moved closer to the corresponding text. And in Fig. 6, should CLU-MEM-RES message be named CLU-MEM-ACK instead (as it is called so in the text)? And it is slightly confusing that in Fig. 4 CLU-ROU-IND is sent from the CH to I/O devices, while in Fig. 6 is it sent in the opposite direction. Maybe one of the messages can be renamed then?
- Algorithm 2 is very well described in the text and thus can be removed. If kept, it should be moved much closer to the corresponding text.

Moreover, there are quite few typos and grammar errors in the paper. Some examples are below:
page 1, left column, line 26: an increase in the number
page 1, left column, lines 39-40: all sensor nodes use limited battery sources
page 1, left column, line 49: Clustering has been proven
page 2, left column, line 36: cannot communicate with each other directly
page 2, left column, line 55: transmit k-bits long packet
page 5, left column, eq. 8: Sk should probably be replaced with Sa?
page 6, left column, line 46: The gateways acted as (remove the) PAN coordinators and
page 6, Table II: There must be a typo in data rate of 11Mbps

[1] D. Chen, M. Nixon, and A. M. Mok, Wireless HART: Real-Time Mesh Network for Industrial Automation: Springer Science+Business Media, 2010.


Reviewer: 2

Comments to the Author
In the article the authors proposed new routing scheme for networks based on IEEE 802.15.4a MAC.
In the paper authors present new algorithm and name it ENERGY-AWARE REAL-TIME ROUTING SCHEME (ERRS).

Description of the algorithm is clear and detailed. Also the state of art is written good.

The results presented in the paper can be interesting for researchers and engineers.

However, the paper has gaps in the research part.

First weakness of the article is focusing only on energy efficiency.
The authors say:"In the current IoT, a larger-scale and complex deployment
has been a noticeable obstacle for minimizing power consumption and routing on real-time."
It would be useful to evaluate algorithms from the point of view of transmission delay and routing on real-time. Make sure readers that the algorithm meets the real time transmission requirements.


Second weakness is connected with simulations and obtained results.

- the results ware obtained only by simulation. (would be useful verification, maybe analytical)

- The authors should describe in detail the collecting results process (how many times authors repeated simulations,
write about the repeatability of the results, write about the results reliability)

- Authors do not write about simulation model verification and validation.

Reviewer: 3

Comments to the Author
A routing scheme that enhances energy consumption and end-to-end delay for large-scale industrial Internet of Things (IoT) systems based on IEEE 802.15.4a MAC has been proposed in this paper. The method is targeted at large-scale systems where data are aggregated through different clusters on their way to the sink. It is meaningful. But this paper has some shortcomings. It must be done by major revision according to the following suggestions:
1) There is no "Related works" Section. In this section, the author should list the current related works on Energy-aware Real-time Routing for Large-scale
Industrial Internet of Things. So many related references must be added.
2) Section "III Energy-aware Real-time Routing Scheme" is shallow, more mathmatics model should be added
3) Explanations should be added for Algorithm 1 and Algorithm 2
4) SIMULATION PARAMETERS in Table II should be corrected.
5) Comparison figures with related methods in Section "IV PERFORMANCE EVALUATION" is not enough, more figures should be added.
6) References 1,13,21 are used to be compared, but these papers are not high levels, so higher references should be added to make comparison to show the advantage of this research.
7) The following current related references are missed, they must be added in Section "References", and give introductions in Section "Related works":
[1] An energy-balanced routing method based on forward-aware factor for Wireless Sensor Network. IEEE Transactions on Industrial Informatics, 2014,10(1):766-773.
[2] A Novel Compressive Sensing Method Based on SVD Sparse Random Measurement Matrix in Wireless Sensor Network. Engineering Computations,2016,33(8):2448 - 2462.
[3] A new constructing approach for a weighted topology of wireless sensor networks based on local-world theory for the Internet of Things (IOT). Computers & Mathematics with Applications, 2012,64(5):1044–1055.
[4] A Novel Approach to Mapped Correlation of ID for RFID Anti-collision. IEEE Transactions on Services Computing, 2014,7(4):741-748.
[5] A Novel Multicast Routing Method with Minimum Transmission for WSN of Cloud Computing Service. Soft Computing, 2015,19(7):1817-1827.
[6] A kind of novel method of service-aware computing for uncertain mobile applications. Mathematical and Computer Modelling, 2013,57(3-4):344-356.
[7] A new medium access control protocol based on perceived data reliability and spatial correlation in wireless sensor network. Computers & Electrical Engineering.2012,38(3):694-702.
[8] Novel Quick Start (QS) Method for Optimization of TCP. Wireless Networks, 2016,22(1):211-222.
[9] New Clustering Routing Method Based on PECE for WSN. EURASIP Journal on Wireless Communications and Networking,2015,2015(162):1-13. DOI: 10.1186/s13638-015-0399-x
8) Polish the whole paper.

Reviewer: 4

Comments to the Author
Energy-efficient and delay-sensitive routing is one of the well-studied research in WSNs as well as IWSNs. Related to the Industrial IoT (IIoT), the paper only considers the hierarchical framework that contains I/O devices, routers, and gateways. These elements are widely used in all traditional networks.

The main drawback of this manuscript is that the technical novelty is very low compared to previous well-studied work. The proposed method is not clear how the energy consumption is reduced. Only hop-count is not enough to validate low-latency transmission. According to my understanding, the paper considers hop-count, cluster formation, and select the lowest consumption path to design the proposed method.

How the power saving constraints are satisfied is not very clear from the manuscript. Although, joint min-hop-count and least energy is used, however, the results are not validated in bounds by experiments or analytical results.

The authors deliberately consider the scheme that does not necessarily consider both real-time and energy-efficient scheme for the comparison. However, there are tremendous amount of work that consider both. Author should compare their results with the state-of-the-art.

The related discussion in performance evaluation section is very weak. No insights are presented.

In Fig. 9, how the distance from sensing layer to the convergence layer affect the performance is well discussed.

Minor Comments:
Two times Communication energy in table II. Please use different notation.
Lots of jargon, e.g., MIMO
In (1), $E_{elec}$ is not properly defined? What do the authors mean by energy consumed to process the data? Some other terms need to be properly discussed.
The optimization condition in (6) is very straight forward.



Reviewer: 5

Comments to the Author
This paper is focus on the routing algorithm which takes energy consumption as a very import factor in IoT based on IEEE 802.15.4a. The proposed algorithm is targeted at large-scale systems where data are aggregated through different clusters on their way to the sink. A hierarchical system framework is employed to promote scalability of IoT elements. By estimating the residual energy and hop counts for each path, it said that the data can be forwarded to the destination through the optimal path by Simulation. It is very good works in the field of Industrial IoT.

A lot of works and reference can be found about IoT routing, in order to highlight the new contribution of this paper, the authors are suggested to test the proposed method and algorithm in a real industry application. At the same time, to compare the proposed algorithms performance with other method if possible.

In addition, some works bout energy-ware in IoT application is suggested to be investigated to highlight the new contribution and new potential application. E.g., the following publication in TII:
Internet of Things in product life-cycle energy management
Journal of Industrial Information Integration 2016
IoT based intelligent perception and access of manufacturing resource towards cloud manufacturing
IEEE Transactions on Industrial Informatics 2014
Internet of Things and BOM based life cycle Assessment of Energy-Saving and Emission-Reduction of Product
IEEE Transactions on Industrial Informatics 10 (2)


Reviewer: 6

Comments to the Author
The paper presents how to model a power consumption of large-scale Wireless Sensors Network where base stations are gateways uploading data to the Internet. The results are interesting and may deserve publication, but revision of text to improve the quality of presentation would be valuable. Please find my comments about the text below. I would recommend to both; proofread the text carefully to make the text clearer and seek assistance from mathematician and engineer to improve the theoretical and experimental aspects.

I. Proofread and revise the text. Pay attention to what is said. For example:
1. The title promises; Energy-aware Real-time Routing for Large-scale Industrial Internet of Things
In fact, the work is devoted mainly to research and discuss the results of modeling an energy-aware routing in the large-scale Wireless Sensors Network, where base stations are gateways uploading data to the Internet.
Both, the proposed model (Section II), as well as the presented energy-aware routing scheme (Section III), and also described performance evaluation (Section IV), does not contain anything, it was specific to the Industrial IoT. If Industrial IoT matters, please take this into account in the system model and the proposed routing algorithm. If not, please consider amending the title, so that it corresponds to the contents of the work.

2. Mathematical formulas in chapter II (System model) and chapter III (Routing scheme) need more attention.
Formula (3), router energy consumption Er fragment; first component Erx(k) according to formula (2) consists of energy for aggregation and compression (see line just below the formula (2)).
So, the k in next component (the sum over indices with z, j, from 1 to M) should be different!
Formula (3), symbols M, O, (see line just below formula (3)) is the number of neighbor nodes of router r. Symbols M, O are different, please specify the differences. The same concerns symbols z, j in second component of Er !
Formula (5), Eround should be the sum of all clusters (not only one Ecluster,i). Moreover, Eg and Er are calculated in clusters, so in formula (5) that fact should be taken into account.
If comments above are valid, in the formula (5) where L>1, it is necessary to preclude the multiple counting of the same components Erj,i
Formula (7), two lines below we read W is the number of I/O devices Whether, it should be added: in the cluster j ? Moreover, so far (formulas (3,4 and text) for determining the number of I/O devices within cluster, the symbol N was used.
Formula (11), how to calculate this? Please explain, how to establish the set {hop counts}. It is crucial element of routing procedure!

3. Routing scheme ERRS in chapter III.
1). Set-up phase.
Page 4, line 51, column left; and position information
Page 4, line 52, column right; residual energy
Page 5, line 37, column left; message containing its hop count
Page 6, lines 49-52, column left; all nodes were in the carrier sensing range
Please specify clearly (e.g. in the chapter system models) ALL assumptions that have been made, and are relevant to the issues under consideration.

2). Fig.4, Updating routing table and calculating transmission power . Does it make sense to calculate routing table for I/O devices (there is only one hop to the cluster head)?
Transmission power, whether residual energy, or energy necessary for communication within cluster? Please explain where these calculated values are later used.

3). Fig.5, CLU-MEM-RES what is it! Whether should be: CRE-CLU-ACK? CLU-ROU-IND on this figure has opposite direction to that on fig.4. Moreover, this direction does not match the description in text (page 4, lines 54-55, column left).

4). Please ensure compatibility of flow diagram at Figure 5, with the one that was shown at Figure 4 and described on page 4. As far as Figure 4 is understandable, the figure 5 requires a major improvement.



II. General and detailed remarks.

1. Please explain the main idea of this contribution, arising from the practical observations or theoretical conclusions. Please provide the rationale of proposed algorithm concerning the aspects described below
- We have energy model (on page 2) where the energy consumed is proportional to distance squared, or even distance raised to the fourth power (see formula (1)). But selecting factor is inversely proportional to distance (not squared or raised to the fourth power).
- Hop counts estimating (see abstract, formula 11, algorithm 2). Where and how it is implemented in the algorithm 2? What amount of energy spends node on hop counts calculation and communication necessary to inform the neighbors of them?
- Potential next-hop. Where ( in the algorithm 2) and how proposed method ensures, that selection of next-hop approaching us to the gateway?
- The proposed approach; favors a small clusters and ignores (see formula (6)) the exchange, between cluster heads, of routing information (what is a pivot activity in real-time routing) during the operation of the network. Small clusters and additional exchange of information increase the number of collisions, and therefore repetition of packages, and increasing energy consumption, but these costs in the objective function (6) are not included.

2. The algorithm 1, page 5, Determine the potential gateway to complete the route; is trivial and may be defined in one sentence: Of the gateways in the neighborhood, select the closest one.
3. The algorithm 2, page 8, Determine the potential next-hop to forward data packets.
- step 8; how to calculate d-rAr i}?
- step 10; brackets {} means set (see steps 11,15), but what does it mean ri.{drri} and ri.{Erri}?
- steps 9-12; During the network lifetime, as time goes on, neighbors who are closer will have less energy than those farther on (because they were more often chosen as a relay) and therefore the condition in Step10 will be increasingly difficult to fulfill. As a result, in spite of that we have neighbors who may be next-hop in routing, the condition in Step10 not met and set NH will be empty.
- steps 14-16; loop FOR i is useless. Step 15 is one operation (see formula (11)), that does not depend on i !
- step 15; where (I didnt see in the algorithm) and how to calculate hop counts ?

4. Page 7, lines 48-49, column right; It is evident that the energy consumption of I/O devices
is reduced significantly by all four routing schemes. Scientifically it sounds bad. What does it mean reduced significantly? Compared to what?
5. Engineering and experimental aspects.
In simulation nodes are distributed randomly (see Fig.7 title). Please determine on figures 8-12, the maximal inaccuracy of presented results. What was the impact of number (ratio) of gateways and routers and I/O sensors, on final results?



Hopefully, that these remarks allow to improve this work.


Reviewer: 7

Comments to the Author
This is a very good manuscript. I can see only two serious issues:

1) The assumption that all nodes can sense the other node carrier (made to avoid the hidden node problem) - it does diminish the realism of the situation. Also, simulation of the wireless communication signal in non-open space (there is no assumption whether this is open space or indoor environment) cannot be realistic due to the multipath problem. The assumptions need to be clarified and their impact on the realisticness of the simulation has to be discussed.

2) The manuscript is too long. I believe that this journal has an eight page limit for manuscripts unless a manuscript is state-of-the-art. This is not state-of-the-art manuscript (only 21 citations). Hence it needs to be significantly shortened.

AE Comments:
Associate Editor
Comments to the Author:
The submitted manuscript is good. The topic is interesting and valid.
The presented topic fits well to the journal scope.
Unfortunately there are some drawbacks need to be eliminated:
- The lack of proper motivation and research problem explanation. The state of the art should be highlighted.
- Research are focusing on simulation only. Some shortcomings exist at all.
- The reference section is not up to date.
- The novelty should be explained and emphasized
- The logical coherence should be improved as well as the whole content including formulas and schemas.
Do you aware how long is your paper? Try to consider if all content is necessary.
After reading the paper, I agree with the reviewers. The paper need to be strongly improved in a number of parts.
For the details see the constituent reviews.